Catalytic Dehydrogenation of Propane and sobutane in Hydrogen Permselective Membrane Reactors John P. Collins' Rakesh Sehgal' Robert W. Schwartz' C. Jeffrey Brinker'v2 Timothy L. Ward* Gary P. Hagen3and Carl A. Udovich3 'Sandia National Laboratories Advanced Materials Laboratory 11 University Blvd. SE AlbuquerqueMM 8716 m S F Center for MicroEngineered Ceramics Farris Engineering Center The University of New Mexico AlbuquerqueNM 87131 3AmocExploration and production Amoco Research Center 15 West Warrenville Rd. Naperville L 6566711 Abstract The dehydrogenation of propane and isobutane was studied in hydrogen permselective packed bed membrane reactors and conventional packed bed reactors. Two different types of developmental membranes were investigated: solgel derived silicabased membranes and a pure palladium thinfilm suppoaed by aporous ceramic substrate. The palladium membranes deactivated and eventually failed when exposed to both isobutane and propane dehydrogenation temperatures above 773 K. Moderate improvements in propylene and isobutylene yields were obtained with the silicabased membrane reactors. An isobutylene yield of 48 mole percent was obtained at a liquid hourly space velocity (LHSV) of 1.8 and temperature of 798 K compared to a yield of 39 percent in a conventional reactor operated with the same flow rate. Similar improvements in propylene yield were obtained when the silicabased membranes were tested in propane dehydrogenation experiments. There was no significant dijjerence in the reaction selectivities for the desired olefin products when the membrane and conventional reactors were operated with the same LHSV. However for a constant value of the olefin yield the membrane reactors had a higher reaction selectivity since the desired yield was achieved at a higher LHSV where there was less time for side products to form. Catalyst deactivation rates were generally greater in the membrane reactors especially when the reactors were operated with high hydrogen removal raes at temperatures of 773 K and above. ntroduction Catalytic dehydrogenation is a potential method for obtaining alkenes for polymerization and other organic syntheses from lowcost saturated carbon feed stocks. Two examples are the selective production of propylene from propane and
DSCLAMER Portions of this document may be illegible in electronic image products. mages are produced from the best available original document.
DSCLAMER This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof nor any of their employees makes any warranty express or implied or assumes any legal liability or responsibility for the accuracy completeness or usefulness of any information apparatus product or process disclosed or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product process or service by trade name trademark manufacturer or otherwise does not necessarily constitute or imply its endorsement recommendation. or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.
isobutylene from isobutane.?hemodynamic limitations on conversion are a significant problem when conventional packed bed reactors are used in both propane and isobutane dehydrogenation. While the propane and isobutane conversion may be increased by operating at higher temperatures this results in an increase in the rate of catalyst deactivation and a decrease in the reaction selectivity for the desired olefin products. Catalytic dehydrogenation in hydrogen permselective membrane reactors may be a potential method for increasing the olefin yield while maintaining an acceptable catalyst deactivation rate and reaction selectivity. There have been several previous studies on both propane and isobutane dehydrogenation in packed bed membrane reactors. Ziaka and coworkers (1993a 1993b) utilized commercially available porous ceramic membranes with Knudsen diffusion permselectivity to study propane dehydrogenation while Sheintuch and Dessau (1996) studied this reaction with commercially available tbickwalled palladiumruthenium and palladiumsilver alloy tubes. Esobutane dehydrogenation has been studied by Matsuda and coworkers (1993) with a supported pure palladium thin film oannides and Gavalas (1993) with a dense silica membrane and Sheintuch and Dessau (1996) with the thickwalled palladium alloy tubes.?he difference between our study and previous studies is that we are investigating the use of porous silicabased membranes with molecular sieving capabilities as opposed to dense membranes or porous membranes with Knudsen diffusion permselectivity. The porous silicabased membranes have a higher hydrogen permeance than the densesilicaand palladium alloy membranes and a higher hydrogen permselectitivy than the Knudsen diffusion membranes. Therefore they are a promising candidate for both isobutane and propane dehydrogenation. n addition we have also performed tests with a supported pure palladium thin film membrane in order to compare the performance of two different types of developmental membranes in the same experimental system. Another objective of the study was to evaluate the membrane reactor performance at space velocities similar to those used in commercial conventional reactors (LHSV = 1 to 2) so that any improvement in performance was obtained at a practical flow rate. Experimental Membrane Preparation Both the silicabased and palladiumceramic membranes consisted of silica or palladium films deposited on the inside surface of Membralox@T17 alumina support tubes obtained from U.S. Filter Corporation. For the palladiumceramic membranes the top layer of the asymmetric support tube was composed of aalumina with a pore diameter of 2 nm. n the case of the silicabased membranes the top layer of the support tube was composed of yalumina with a pore diameter of 5 nm. For ease of handling and testing the membrane supports (supplied by U.S. Filter in lengths of 25 and 75 cm) were cut into 6 cm sections using a diamond wafering saw..
Electroless plating was used to fabricate the palladium membranes. The plating conditions for membranes tested in propane dehydrogenation experiments were similar to those described by Collins and Way (1993). For the membrane tested in isobutane dehydrogenation experiments the electroless plating/osmotic pressure procedure developed by Yeung and Varma (1995) was used to deposit the palladium film. n this process the outside of the membrane support tube was immersed in a 5.8 Molal sucrose bath at 6 C while plating solution was pumped through the inside of the tube. A polymeric silica sol was used to fabricate the silicabased membranes. The silica films were deposited on the support tubes by a solgel process that combines the features of slip casting and dip coating. The supports were calcined at 823 or 873 K following deposition of the silica film. Additional information regarding the sol preparation and membrane coating procedures is presented by Sehgal and coworkers (1994) and Sehgal(l996). Membrane and Conventional Reactor Experiments Propane and isobutane dehydrogenation experiments were conducted in packed bed membrane reactors and conventional packed bed reactors operated under the same conditions. The conventional reactor design was similar to the membrane reactot design except the membrane was replad by a quartz tube with the same inside diameter (.7 cm).the reactors were packed with a platinumloaded aluminosilicate molecular sieve catalyst developed by Amoco Corporation. Catalyst preparation procedures were similar to those described by Kaminsky and coworkers (1993). The experimental system was similar to the one used by Collins and Way (1993 1994). A shell and tube type design consisting of a single membrane tube inside a quartz tube was used for the membrane module. The ends of the membrane were connected to nonporous alumina tubes using Swaglok@compression fittings with seals (ferrules) made from GEUFOL@ribbon. Membrane reactor experiments were conducted at temperatures ranging from 723 to 798 K for isobutane dehydrogenation and 773 to 848 K for propane dehydrogenation. The reactor feed composition was either pure isobutane or propane or mixtures of isobutane or propane with hydrogen. The total pressure on the reaction side was approximately 1 atmosphere with a sweep side pressure of approximately.9 atmospheres. (Note that atmospheric pressure in Albuquerque New Mexico is approximately.84 atmospheres). A nitrogen sweep gas was used to obtain the partial pressure driving force for hydrogen permeation. The ratio of the inlet sweep gas to inlet reactor feed gas flow rates (FR) was varied from 2 to 1. Additional information regarding the experimental conditions and procedures is presented by Collins and coworkers (1996). Catalyst deactivation occurred to varying degrees in both the membrane and conventional reactor experiments. The catalyst was reactivated by passing air through the system at 723 K or 748 K. Following catalyst reactivation the same
time schedule was followed for the membrane and conventional reactor experiments in order to compare their performa~ce. Results and Discussion Results of Gas Perkation Experiments Gas permeation experiments were performed during the course of the membrane reactor studies to evaluate the membrane performance. Prior to initiating the reactor experiments pure gas permeation experiments were conducted with hydrogen and nitrogen at 723 K or 773 K. Results of these experiments are summarized in Table 1. Table 1 Results of nitial Pure Gas Permeation Experiments Pd #12 Pd #22 Silica #1 Silica #2.24 77.71 192.16 1.2.31 12.8 Silica #3.32 18.8 Silica ##4 Silica #5.26 15.2.13 11.6 Units for the hydrogen permeance (q) are cm3(stp)/s*cm2*cmhg. Hydrogen permeance values for the palladium membranes are listed in units of cm3 (STP)/s*cm cm Hg instead of cm3(stp)/s*cm *cm Hg so that they may be compared to the silica membranes. The listed hydrogen permeance values for the palladium membranes are applicable for a tube side hydrogen pressure of 115 cm Hg and permeate side hydrogen pressure of 64cm Hg. The palladium membranes had the best initial perfonnance because their hydrogen permeance and hydmgenhitrogen permelectivity were higher than those obtained for the silicabased membranes. The small amount of hydrogen permeation observed is attributable to defects in the palladium film andor leakage through the GRAFOL@seals used in the membrane module. Palladium Membrane #1 was prepared using the standard electrolessplating procedure while Palladium Membrane #2 was prepared with the electroless plating/osmotic pressure procedure. Palladium film thicknesses for the two membranes were approximately 12 pm for Palladium Membrane #1 and 7.5 p.m for Palladium Membrane #2. Nitrogen permeances for the two membranes were similar which indicates that the application of osmotic pressurereduces the film thickness requited to fabricate palladium membranes with high hydrogen pennselectivities. Although the hydrogednitrogen permselectivities observed for the silicabased membranes were lower than those obtained for the palladium membranes the selectivities were significantly higher than the Knudsen diffusion value of 3.74 which indicates Qat gas permeation was occurring via the molecular sieving transport mechanism.
Hydrogen permeation rates were also monitored during the membrane reactor experiments to evaluate membrane performance under reaction conditions. Hydrogen permeation rates measured during isobutane dehydrogenationexperiments conducted at 45 C over an 8 hour time period with Palladium Membrane #2 and Silica Membrane #4 are shown in Figure 1. As shown in Figure 1 the palladium membmne deactivated when it was exposed to reaction conditions. The hydrogen permeation rate decreased significantly with time on s t r m while at the same time the hydrogen partial pressure in the effluent (residue) stream from reaction side increased. Therefore hydrogen was still being generated on the reaction side but it was not being removed through the palladium membrane. t was possible to reactivate the palladium membrane by flowing hydrogen through the system after the membrane reactor experiments were completed. The hydrogen permeance was generally restored to about half of its original value by flowing hydrogen through the system overnight. The original permeance could then be restored by passing air through the system for 5 minutes at 723 K. ts n 3 1 " ' l " ' l " ' \ Q) LHSVFR= =2.54 4.5 z N 25 W a =s 2 8 Q) h u 15 c 1 2 4 6 8 1 Time on Stream (hours) Yield A Pd MR H Permeation Rate [3 P d MR 8 Silica MR Yield Figure 1 +Silica 2 MR HZ permeation Rate sobutylene Yields and Hydrogen Permeation Rates Measured in sobutane Dehydrogenation Experiments Conducted with Silica Membrane #4 and Palladium Membrane #2 at 723 K.
The palladium membranes were also tested for propane and isobutane dehydrogenation at temperatures of 773 K and above. At 773 K the hydrogen permeation rates decreased with time on stream but permeation rates of the other gases were approximately constant. When the temperature was increased to 798 K hydrogen permeation rates continued to decrease but permeation rates of the other gases lncreased with time. This increase in gas permeation rates occurred because defects developed in the palladium film when the membranes were exposed to reaction temperatures above 773 K. Therefore it appears that a temperaturerange of 773 to 798 K is the upper limit when palladium membranes fabricated using the conditions employed in this study are used for either isobutane dehydrogenation or propane dehydrogenation. Although the hydrogen permeance of the silicabased membranes was relatively constant when exposed to reaction conditions c o m p d to the palladium membranes the permeance of all of the silicabased membranes tested in this study decreased over time. For example the hydrogen permeance of Silica Membrane #2 degeased by 34 percent after 1 hours of testing while the hydrogednitrogen permselectivity remained approximately constant. Similar behavior was observed with the other silicabased membranes. The drop in hydrogen permeance does not result from deactivatiodcoking due to exposure to reaction conditions because the hydrogen permeance was not recovemi when the catalyst was reactivated. Alternative fabrication procedures designed to improve the stability of the silicabased membranes are under investigation in our laboratories. Results of Membrane Reactor Experiments sobutylene yields obtained in the membrane and conventional reactor experiments conducted at 723 K are shown in Figure 1 along with the observed hydrogen permeation rates. The isobutylene yields for the palladium membrane decreased with time while those for the silicabased membrane were relatively constant which indicates that membrane deactivation rather than catalyst deactivation is responsible for the reduction in palladium membrane reactor yields at this temperature. n their study of isobutane dehydrogenation in a palladiumceramic membrane reactor Matsuda and coworkers (1993) concluded that catalyst deactivation rather than membrane deactivation (hydrogen permeance) was the limiting factor. There are several differences between the present study and the study performed by Marsuda and coworker s which may account for the discrepancy in results and conclusions. First the catalysts used in the two studies were different so catalyst deactivation rates are expected to be different. The space velocities used by Matsuda and coworkers were also approximately 5 times lower than those reported in Figure 1 and the hydrogen permeation rate is more sensitive to changes in hydrogen permeance when higher space velocities are used. n addition Matsuda and coworkers only performed experiments over a one hour time period. As shown in Figure 1 the hydrogen permeance continues to decrease with time on stream after one hour. Finally the procedure used by Matsuda and coworkers to determine hydrogen permeances following exposure to reaction conditions is not clear. n our..
study we were not able to obtain a value for the hydrogen permeance directly following the conclusion of membrane reactor experiments because it increased as soon as the membrane was exposed to hydrogen. sobutylene yields and isobutylene reaction selectivities obtained with Silica Membrane #5 are shown in Figure 2 as a function of the LHSV. An isobutylene yield of 56 mole percent was obtained in the membrane reactor at an LHSV of.9 (catalyst volume = 2 cm3 catalyst wt. = 1 g inlet isobutane flow rate = 6.9 cm3 (STP)/min) compared to a yield of approximately 4 percent in a conventional reactor o p t e d with the same LHSV. The dif aence between membrane and conventional reactor yieids decreased as the LHSV increased. At the more practical LHSV of 1.8 the membrane reactor yield was only 48 mole percent compared to a yield of 39 percent in the conventional reactor. Figure 2 sobutyiene Yield and Reaction Selectivity Versus LHSV For Dehydrogenation Experiments Conducted at 798 K with Silica Membrane #5 and Conventional Reactor. As shown in Figure 2 there was no significant difference in the isobutylene reaction selectivities between the membrane and conventional reactors at any particular LHSV. However for a constant value of the isobutylene yield the membrane reactor had a higher isobutylene reaction' selectivity since the desired yield was achieved at a higher LHSV where there was less residence time for undesired side products to form. For example an isobutylene yield of 42 mole percent and
reaction selectivity of 97 mole percent was achieved in the membrane reactor at an LHSV of 2.75 compared to a yield of 4.5 mole percent and reaction selectivity of 93 mole percent in the conventional reactor operated with an LHSV of.9. Therefore a membrane reactor can achieve a desired value of the isobutylene yield with a smaller reactor volume and higher reaction selectivity than a conventional reactor. sobutylene yields obtained in experiments conducmi at 773 K with Silica Membranes #4 and #5 are shown in Figure 3 as a function of time on stream. Due to its greater hydrogen permeance the isobutylene yields for Silica Membrane #4 were expected to be higher than those for Silica Membrane #5. However the yields for Silica Membrane #5 were unexpectedly higher than those obtained with Silica Membrane #4 because catalyst deactivation was a significant problem when Silica Membrane #4 was tested at this temperature. Apparently the greater hydrogen removal obtained with this membrane was offset by the increased rate of catalyst deactivation which resulted from the &ad hydrogen concentration in the catalyst bed. Therefore for any given catalyst the best membrane is not necessarily the membrane with the greatest hydrogen permeance. The effect of hydrogen removal on catalyst deactivation rates must also be considered. Sheintuch and Dessau (1996) have also noted that isobutylene yields in experiments conducted with a palladiumruthenium alloy tube passed through a maximum with increasing hydrogen concentration in the catalyst bed. n 5 m h Q) 2 45 5 4 1 W 2 "h 35 c s 3 Figure 3 l l A ' l l LHSV =.91 FR = 1 Pure i C p l a reactor feed A A H s ' = Q) a Silica #5 Silica #4 Conv. Reactor # A # # t r A A n 2 \ 4 6 8 Time on Stream (hours) ~ t 1 sobutylene Yield ' Versus Time on Stream For Dehydrogenation Experiments Conducted at 773 K with Silica Membranes #4 and #5..
Propane dehydrogenation experiments were also performed with Silica Membranes #1 2 and 3. Detailed information regarding the results of these experiments is presented by Collins and coworkers (1996). Results of the propane dehydrogenation study were similar to those obtained in isobutane dehyhgenation in that only moderate improvements in propylene yields were obtained when the membrane reactors were operated at a practical LHSV. Catalyst deactivation was a significant problem because higher temperatures than those employed in the isobutane dehydrogenation study were required to obtain reasonable propylene yields. mprovements in the hydrogen permselectivity of the silicabased membranes are needed in order to reduce the amount of isobutane and propane that permeate through the membrane and reduce back permeation of the sweep gas. Permeation losses of the reactants ranged from approximately 5 to 2 percent depending on the specific membrane tested and the LHSV. The permeation of reactants reduces the yield to olefins since the dehydrogenation reactions do not occur on the permeate side. Nitrogen concentrations in the outlet (residue) stream from the membrane reactor ranged from approximately 5 to 35 mole percent which would create a downstream separations problem if improvements in the hydrogen permselectivity are not made. Conclusions Pure palladium thin films are not a promising membrane candidate for either isobutane or propane dehydrogenation membrane reactors because the palladium deactivates when exposed to reaction conditions. Future research should focus on the identification and preparation of thin film palladium alloy materials which have a greater resistance to deactivation than pure pauadium and sufficient hydrogen permeance. Silicabased membranes appear to be a promising candidate for these reactions provided their long term stability can be improved and their hydrogen permselectivity can be increased without a significant reduction in hydrogen permeance. mprovements in catalyst stability are also required in order to operate the catalyst bed under the lower hydrogen and higher olefin concentrations of the membrane reactor with acceptable catalyst deactivation rates. Acknowledgment The portion of this work performed at Sandia National Laboratories was supported by the U.S. Department of Energy under Contract # DEACO494AL85OOO. Rakesh Sehgal was supported by funding from the Electric Power Research nstitute and the Gas Research nstitute. References Collins J.P.; Sehgal R.; Schwartz R.W.; Brinker C.J.; Ward T.L.; Hagen G.P.; Udovich C.A. Catalytic Dehydrogenation of Propane in a Hydrogen Permselective Membrane Reactor. Submitted to d Eng. Chem. Res. Mar& 1996.
Collins J.P.; Way J.D. Preparation and Characterization of a Composite PalladiumCeramic Membrane. Znd. Eng. Chem. Res. 1993 32 36313. Collins J.P.; Way J.D. Catalytic Decomposition of Ammonia in a Membrane Reactor. J. Mernbr. Sci. 1994 96 259274. oannides T; Gavalas G.R. Catalytic sobutane Dehydrogenation in a Dense Silica Membrane Reactor. J. Membr. Sci. 1993 77(2&3). 2722. Kaminsky M.P.; Froment G.R.A.; Dehertog W.J.H. Catalyst and Process for Hydrocarbon Dehydrogenation. nternational Patent Application #PCTNS93/1273 February 1993. Matsuda T.; Koike.; Kubo N; Kiuchi E. Dehydrogenation of sobutane to sobutene in a Palladium Membrane Reactor. Applied Catalysis 1993 96(1) 313. R. Sehgal Microporous Gas Separation Membranes: Fundamentals 'reparalion and CharacterizationPhD Thesis The University of New Mexico 1996. Sehgal R.; Huling J.C.; Brinker C.J. Use of Silica Sols in norganic Moecular Sieving Membranes. Proceedings of the Third nternational Conference on norganic Membranes July 19948594. Sheintuch M; Dessau R.M. Observations Modeling and Optimization of Yield Selectivity and Activity During Dehydrogenation of sobutane and Propane in a PD Membrane Reactor. Chem Eng. Sci. 1996 51(4) 523547. Yeung K.L.; Vanna A. Novel Preparation Techniques for Thin MetalCeramic Composite Membranes. ACheJ 1995 41(9) 21312139. Ziaka Z.D.; Minet R.G.; Tsotsis T.T. Propane Dehydrogenation in a Packed Bed Membrane Reactor. ACheJ. 1993a 39 526. Ziaka Z.D.; Minet R.G.; Tsotsis T.T. A High Temperamre Catalytic Membrane Reactor for Propane Dehydrogenation. J. Membrane Sci. 1993b 77221232.